Rethinking Seizure Care Blog

Can an Algorithm Predict Seizure Origin in Pediatric Patients?

Posted by RSC Diagnostics on Aug 1, 2019

Blue Cogwheel Gear Mechanism Icon on Dark Digital Background.-1

A new and computational approach to surgery developed at Boston Children’s Hospital may better facilitate surgery in children with epilepsy who do not respond to anti-seizure medications. This approach, described in the journal Neurosurgery, has the potential to simplify the way doctors monitor seizures for surgical pre-planning in children with epilepsy. 

How is the Origin of a Seizure Determined?

Currently, and depending on the patient, it may be necessary to perform an initial invasive surgery to discover the exact regions of the brain where a patient’s seizures originate before performing a second surgery to remove the affected brain tissue. In the initial surgery, a network of electrodes is placed on the surface of the patient’s brain. Then, the patient undergoes a long-term EEG monitoring study, typically a week, while doctors wait for a seizure to occur. Once the origin of seizures is pinpointed, the patient will have a second surgery to remove the afflicted areas. 

Is it Possible to Determine the Origin of a Seizure without Observing One?

A new approach developed by Director of Epilepsy Surgery at Boston Children’s Hospital, Dr. Joseph Madsen, and Computational Biophysicist, Dr. Hyoung Park, has the potential to be performed in a single surgical session, without the need to observe a seizure. This approach is potentially cost-saving (one surgery vs. two) and could reduce a patient’s risk. However, how can this new approach remove the need to observe a patient’s seizure?

Could an Algorithm Help Predict the Origin of a Patient’s Seizures?

Dr. Madsen states, "So rather than wait for the patient to have a seizure, we set out to find patterns of interaction between various points in the brain that might predict where seizures would eventually start."

These patterns of interaction are revealed by an algorithm known as Granger causality analysis. Originally developed for use in the financial market, this statistical algorithm uses probability to derive conclusions. In this study, the doctors analyzed 25 randomly selected patients who had previously had long-term EEG monitoring studies. Madsen and Park analyzed their interictal EEG results (data obtained in the interval between seizures) and generated a map of those patients’ brain networks. They superimposed those mappings over an image of the brain to predict seizure origin. Interestingly, the brain regions predicted to be the cause of seizures strongly correlated to the brain regions that actually caused seizures.

What’s Next?

After the study was completed, the results were reviewed by ten board-certified epileptologists for verification that mapping has the ability to predict the regions of the brain that cause the seizures. A strong causality was found. Dr. Madsen states that although the results are promising, further validation and refinement are needed before the Granger causality analysis can be used clinically. The next step is to evaluate how the algorithm can be used by neurophysiology when evaluating EEG readings.




Topics: Pediatric Epilepsy